ArghDA M10: Mizar adapter — heavy tail complete (a real verify, not detect-only)#46
Merged
Merged
Conversation
…etect-only) The "highest uncertainty" backend turned out tractable: Mizar 8.1.15's statically-linked i386 verifier runs on x86_64 with no 32-bit libs, so M10 is a genuine dogfooded verify rather than the planned detect-only stub. arghda is now a nine-backend prover-parametric workspace; the heavy tail (M8-M10) is complete. Adapter (src/prover/mizar.rs, `Mizar`): - Assistant kind, extensions [miz], command `verifier`. - No lint pack, by design: Mizar has no sorry/admit escape hatch — an incomplete proof is just a verifier error, and a normal article cannot introduce axioms — so soundness IS the verifier and the verdict is binary. - check_file, ground-truthed vs Mizar 8.1.15: `MIZFILES` unset → Unavailable (the MML data dir is the precondition — honest, never a fabricated pass); binary absent → Unavailable; otherwise copy the .miz (+ siblings) to a temp dir, run `accom` then `verifier` (env MIZFILES), and read the sibling `<stem>.err` — EMPTY + exit 0 → Proven, non-empty / exit≠0 → Error (Mizar reports errors in .err, the authoritative signal). - Imports: the `environ` block's directives (vocabularies/notations/theorems/…), lower-cased to match on-disk stems; MML references are external and dropped. - Roots: none — Mizar has no aggregator convention (honest empty set). - probe() overridden to `verifier -v`. Provisioning at source: install_mizar upgraded from a warn-stub to a real installer (static i386 tarball from mizar.uwb.edu.pl → /opt/mizar + symlinks), with a MIZAR_TAR pin and a doctor verify line. Wired: backend_for "mizar", KNOWN_BACKENDS, all four `--backend` help strings, the CLI about line. Verification (actually run): cargo fmt --check, clippy -D warnings, cargo test (167 pass: +6 mizar unit, +2 mizar integration; graph tests hermetic — environ text only, run tool-less), check-spdx.sh — all green. Dogfooded vs real Mizar 8.1.15 (MIZFILES=/opt/mizar): `check` → good=proven (theorem for X being set holds X = X, importing XBOOLE_0) / broken=error / empty=proven; `doctor` → [OK] mizar assistant Mizar Ver. 8.1.15. Remaining (M10 at 85%): local-article cross-references need prel export (miz2prel) for the throwaway session; no root convention. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012MpYSh6Wy8YMBH2E3qVyT7
hyperpolymath
marked this pull request as ready for review
July 1, 2026 13:15
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
The third and last heavy-tail backend — and the "highest uncertainty" one. It turned out tractable: Mizar 8.1.15's statically-linked i386 verifier runs on x86_64 with no 32-bit libs, so M10 is a genuine dogfooded verify, not the planned detect-only stub.
arghda is now a nine-backend prover-parametric workspace (agda, agda-cubical, idris2, lean4, coq, isabelle, mizar as Assistants; z3, cvc5 as Solvers). The heavy tail (M8–M10) is complete.
How
Adapter —
src/prover/mizar.rs(Mizar)[miz], commandverifier.sorry/admitescape hatch — an incomplete proof is simply a verifier error, and a normal article cannot introduce axioms — so soundness is the verifier and the verdict is binary.check_file, ground-truthed vs Mizar 8.1.15:MIZFILESunset →Unavailable(the MML data dir is the precondition — honest, never a fabricated pass); binary absent →Unavailable; otherwise copy the.miz(+ siblings) to a temp dir, runaccomthenverifier(envMIZFILES), and read the sibling<stem>.err— empty + exit 0 →Proven, non-empty / exit ≠ 0 →Error(Mizar reports errors in.err, the authoritative signal).environblock's directives (vocabularies/notations/theorems/…), lower-cased to match on-disk stems; MML references are external and dropped.probe()overridden toverifier -v.Provisioning at source:
install_mizarupgraded from a warn-stub to a real installer (static i386 tarball frommizar.uwb.edu.pl→/opt/mizar+ symlinks), with aMIZAR_TARpin and adoctorverify line.Wired:
backend_for "mizar",KNOWN_BACKENDS, all four--backendhelp strings, the CLI about line.Verification (actually run)
cargo fmt --check— cleancargo clippy --all-targets -- -D warnings— cleancargo test— 167 pass (+6 mizar unit, +2 mizar integration; the graph tests are hermetic —environtext only, run tool-less)bash scripts/check-spdx.sh .— OKDogfooded against real Mizar 8.1.15 (
MIZFILES=/opt/mizar):arghda check --backend mizargood.miz(theorem for X being set holds X = X, importing XBOOLE_0)provenbroken.miz(undefined reference)errorempty.miz(empty environ+begin)provenarghda doctor→[OK] mizar assistant Mizar Ver. 8.1.15Scope
M10 at 85%. Remaining: local-article cross-references need prel export (
miz2prel) for the throwaway session; no root convention.🤖 Generated with Claude Code
https://claude.ai/code/session_012MpYSh6Wy8YMBH2E3qVyT7
Generated by Claude Code